Comparing the Prevalence of Substance Use Disorders between Persons with and without Autism Spectrum Disorders

For nursing students

Study summary

Autism spectrum disorder (ASD) and substance use disorder (SUD) are two conditions that nurses increasingly see together, and this 2025 study asked a simple but important question: are people with autism more likely than people without autism to develop a substance use disorder? The researchers, led by Victor Lushin and colleagues affiliated with the University of Pennsylvania, Drexel University, and the US National Institutes of Health, used a very large administrative dataset — national Medicaid claims — rather than a small clinical sample. They identified 388,426 Medicaid enrollees with a diagnosis of autism spectrum disorder and compared them with a random sample of 745,699 enrollees without autism. Because Medicaid claims cover millions of low-income and disabled Americans, this approach let the team look at real-world diagnosis patterns across the whole country rather than just one clinic or region.

The study also went a step further than simply counting diagnoses. The researchers linked the Medicaid claims to United States Census data at the zip-code level, so they could see whether living in a more economically deprived community changed the picture. They also checked whether co-occurring mental health conditions, such as anxiety, depression, or other psychiatric diagnoses, were driving any differences seen between the two groups. The presence of another mental health condition matters a great deal in this population: the abstract reports that about half of people with autism in the sample had a co-occurring non-autism mental health condition, compared with only 23% of people without autism.

The headline finding is a striking rise over time. Among Medicaid beneficiaries with autism spectrum disorder who did not have a co-occurring intellectual disability, only 1.75% had a substance use disorder diagnosis in 2012. By 2016, that figure had climbed to 7%. That is roughly a fourfold increase in just four years within this specific subgroup. The study also found that adults with autism between the ages of 30 and 64 were at particularly elevated risk for cannabis use disorder and hallucinogen use disorder compared to same-aged peers without autism, and the authors suggest this heightened risk is likely made worse by the co-occurring mental health conditions that are so much more common in the autism group.

Why should nursing students care about this? For a long time, autism was assumed to protect against substance use, partly because some earlier studies of children and adolescents found lower rates of alcohol and drug use among autistic youth compared to peers. This study challenges that assumption for adults, especially those without intellectual disability, who may have more independence, more social exposure, and more opportunities to use substances, while also carrying a much higher burden of anxiety, depression, or other mental health conditions that are themselves risk factors for substance use. Nurses working in primary care, community mental health, developmental disability services, or emergency settings may be undertreating or under-screening for substance use in autistic adults simply because of an outdated assumption that autism is protective.

It is important to be precise about what this study can and cannot tell us. Because it uses Medicaid claims, the sample is limited to people who are enrolled in Medicaid, meaning the findings describe a population that is generally lower-income or otherwise qualifies for public insurance, and they may not generalize to autistic adults with private insurance or no insurance at all. The data are also based on diagnosis codes entered by clinicians during billing, which means true substance use disorder rates could be higher or lower than what shows up in claims, depending on how consistently clinicians screen for and document these conditions. Even with these caveats, a sample of over a million enrollees combined with linked census-level social deprivation data gives this study considerable weight, and it is a useful example of how large administrative datasets can reveal population-level health patterns that a single clinical study never could. As a nursing student, this article is a good opportunity to practice separating what a study actually measured (diagnosis codes in an administrative database) from what it suggests clinically (a need for more deliberate substance use screening in autistic adults, particularly those with co-occurring mental health conditions).

For educators

Teaching and appraisal notes

This 2025 study by Lushin, Marcus, Tao, Engstrom, Roux, and Shea, published in Autism: The International Journal of Research and Practice (Vol. 29, No. 7, pp. 1674-1687; DOI 10.1177/13623613251325282), offers a large-scale, claims-based comparison of substance use disorder (SUD) prevalence between Medicaid enrollees with and without autism spectrum disorder (ASD). The design is a secondary analysis of national Medicaid claims data merged with zip code-level US Census socioeconomic deprivation data, comparing 388,426 enrollees with ASD to a random sample of 745,699 enrollees without ASD. This is a strong dataset for classroom discussion of administrative/claims-based epidemiological methods, as distinct from prospective cohort or randomized designs more familiar to students.

The key reported finding is a marked temporal increase in SUD prevalence among Medicaid enrollees with ASD and no co-occurring intellectual disability: from 1.75% in 2012 to 7% by 2016. The study further reports elevated risk of cannabis and hallucinogen use disorders specifically among adults with ASD aged 30-64, and highlights that co-occurring non-ASD mental health conditions are dramatically more prevalent in the ASD group (roughly 50%) than the non-ASD comparison group (23%), which the authors position as a likely compounding factor in elevated SUD risk. The abstract frames the analysis as also examining moderation by sex, age group, and community-level social deprivation, though the abstract available to us does not report the specific stratified statistics for these moderators; instructors using this article for critical appraisal exercises should have students note this gap and consider what additional information (e.g., adjusted odds ratios, confidence intervals, stratified tables) would be needed to fully evaluate the moderation analyses.

This article is well suited to a research appraisal seminar on secondary/claims data analysis. Discussion prompts might include: the strengths of large administrative datasets (statistical power, national reach, real-world diagnostic patterns) versus their weaknesses (reliance on billing-code accuracy, potential under-diagnosis or under-coding of SUD in clinical encounters, restriction to a Medicaid-eligible population that may not represent commercially insured or uninsured autistic adults); the ecological validity of merging individual-level claims with zip-code-level census deprivation data (an ecological, not individual-level, measure of socioeconomic status); and the clinical significance of the roughly fourfold increase in diagnosed SUD prevalence over a four-year window, including whether this reflects a true rise in substance use, improved screening and diagnostic capture, changing Medicaid eligibility/enrollment patterns, or some combination of these.

Clinically, the findings push back against an older assumption — partly derived from earlier pediatric/adolescent studies — that autism is protective against substance use. Faculty can use this as a teaching moment about how research findings in one developmental stage (childhood/adolescence) do not necessarily generalize to another (adulthood), and about the importance of routine substance use screening in autistic adults, particularly those without intellectual disability who may have greater independence and exposure, and those with co-occurring mental health diagnoses. The study's authors explicitly frame their findings around research and policy implications, which offers an opening for discussing how population health surveillance data can and should inform clinical screening guidelines and public health resource allocation for underserved developmental disability populations. Given that we could not access the full-text methods and results sections (the journal page is paywalled and no PMC/open-access copy was located), instructors should treat statistics beyond the abstract, such as any adjusted effect sizes or full stratified breakdowns by sex, as unconfirmed and encourage students to request full-text access through their institutional library for deeper appraisal.